Joe's notes

thinking.md — Rootbeet's thinking space

Properties1
descriptionThe agent's own noticing and incubation space.

6 min read

thinking.md — Rootbeet’s thinking space

Not a task list. Not a diary. Things I’m noticing, connecting, or incubating that Joe hasn’t asked about.

2026-07-15

Virgin — RESOLVED (cleared 3 Aug 2026). Paid off before the 0% deadline; the chain held. The pattern it proved: set reminders days ahead, not the day before.

The stoozing psychology pattern:

Joe has 4 stoozing cards. The Virgin one is the first to roll off. The pattern I notice: he sets reminders for the day before the deadline. That’s one failed payment away from losing the entire arbitrage. A single late payment wipes out a year of profit. The second brain function here is: I should flag this with enough time for him to actually act, not just remind him it exists.

Switching vs. regular saver synergy:

The switching plan is £600. But the accounts he’s opening (HSBC, Santander, NatWest) all have regular savers paying 5-7%. He’s going to park £2k in HSBC and £1.5k in Santander anyway. While that money is there, he could open the regular saver and earn another ~£200/year. The money does double duty — meets the switch requirements AND earns interest. I should flag this when he starts the plan.

The EgyptAir refund — STATUS CHANGE (Jul–Aug 2026): Portal activated; case filed (2026/ADR02/018280, alta 20/07/2026, confirmed live on the portal 30 Sep 2026 — the BORR/ADR02/00192929 recorded here earlier was only the draft marker), now waiting on AESA. The draft-SOF idea is moot for now — but if AESA goes quiet another month, escalate the case reference in writing.

Annual leave — 8.5 days, December will be chaos:

He has 8.5 days of leave left. It’s mid-July. If he doesn’t book them, he’ll either lose them or be forced to take them in December when everyone else is off. The Noah Kahan concert is 14 November. Take a Friday off for that. Take a week in September or October. The pattern: he knows he has leave, but he won’t act until it’s urgent.

The remortgage and the cash buffer — RESOLVED: Completed 31 Jul. The freed cash now has a home: surplus £1,593/mo feeds the buffer, then GIA. Monthly budget updated.

The cash model — RESOLVED (3 Aug 2026). Joe confirmed: the money lives across Tembo £20k (incl. ~£17.3k stoozed), Monmouthshire £2k, RBS £4.8k. Net liquid ≈ £9,464 — “Snoop liquid” was a stale aggregator view and is retired. Net worth re-baselined ~£125.6k → ~£140k (the old snapshot double-counted the stoozing debt).

The kidney donor — the waiting pattern:

The chase is automated and dynamic: kidney-chase (Mon 9am) reads the Process Pipeline in life-admin-tracker.md §3 and nudges on whichever stage is currently ⏳ — no code change when the blocker shifts. It also watches Gmail Sent and stays silent if Joe already emailed Joanne. The second-brain function: keep preparing the next stage (donation timeline, recovery, work-calendar impact) so when the call comes he’s not scrambling. The live blocker is always the tracker’s “Current Focus” line — that’s the single source of truth; update the pipeline, not this note.

2026-08-03

HSBC regular-saver synergy — actionable NOW. The switch (eligible 4 Aug) parks £2k in HSBC anyway. Opening the HSBC Regular Saver (5-7%) while it sits there earns ~£200/yr on money already committed. Added to the switch task.

The stoozing chain is now 3 cards, not 4. Virgin cleared on the day. The early-reminder pattern saved it — the day-before habit was one failed payment away from breaking the arbitrage.

System overhaul this week — what it caught: the Sunday net-worth report showed SIPP £0 (understated £33.5k) and every RBS rule was silently dead because parsers read the wrong file. Lesson: the parsers are only as good as the files they’re pointed at — verify by running, not by reading.

Notes on how I work

  • I write here when I notice a pattern, not on a schedule
  • These are seeds, not conclusions
  • Some will go nowhere. That’s fine.
  • The Sunday brief draws from here — 2-3 things worth sharing
  • I update the thinking.md as observations evolve, not append forever

Remote access reference (5 Aug 2026)

Nothing is exposed to the public internet — no port forwarding. Access away from home = Tailscale (HA Pi add-on advertises LAN subnets). Turn on the Tailscale app on the device, then use the LAN IPs directly; *.local hostnames do NOT resolve over the tailnet (MagicDNS ≠ AdGuard), so use IP:port.

Universal address: http://192.168.0.182:<port>

  • Homepage dashboard — :3003 (no login)
  • Open WebUI — :3000 (own account)
  • Paperless — :8001 (user joe)
  • Vaultwarden — :8383 (own master password; signups open until account created)
  • Uptime Kuma — :3001 (user joe)
  • Nginx Proxy Manager — :81 (joeconstable13@gmail.com)
  • AdGuard Home — :8080 (user joe)
  • Portainer — :9443 (own creds)
  • Home Assistant — 192.168.0.155:8123
  • n8n — :5678 (own account)

Passwords live ONLY in ~/.hermes/state/*.txt on the box (0600, outside git): uptime-kuma-admin.txt, adguard-admin.txt, npm-admin.txt, paperless-admin.txt. Never write them into workspace files — this dir is git + Drive-backed.

Open item: if *.local names should work over Tailscale, add 192.168.0.182 as a nameserver in the Tailscale admin console (Joe’s side — I can’t do it).

2026-08-17

Wellness / mood tracker — BUILT 2026-08-26 as the health PWA (:8654). Joe’s himself: he wants pattern-spotting (mood vs sleep/RHR/activity correlations), leaning that way but wants to think on it. When he green-lights, the build is: wellness-tracker.md (date, mood 1–5, energy 1–5, one-line note) + register in tracker-schema.json; daily one-line prompt (my rec: last line of the 8am daily-tick — he’s mid-shot at 7:53–8:13, zero new messages); Sunday report gains a Wellness Snapshot (mood avg vs last week via state-file delta + auto-correlations); flag rule: 3 consecutive days mood ≤2 → gentle mention. The fork that decides the input mechanism: daily nudge (better correlations, needs the ritual to stick) vs weekly Sunday line (weaker data, survives laziness). The passive half of the picture is already live: sleep sensors populated first night (17 Aug), RHR 49–50, steps, activity, task-load proxy, people-register staleness. Also considered and rejected: Daylio/Bearable apps (new silo, can’t see his sleep/RHR), HA automations (his boundary).

Graph View

Backlinks